A free AI visibility audit checklist you can run this week
TL;DR An AI visibility audit checks whether ChatGPT, Perplexity, Google AI Overviews, Claude and Gemini name or cite your brand for the questions buyers already ask. Run a fixed prompt panel, record who gets named, fix
TL;DR
An AI visibility audit checks whether ChatGPT, Perplexity, Google AI Overviews, Claude and Gemini name or cite your brand for the questions buyers already ask. Run a fixed prompt panel, record who gets named, fix crawl and content gaps, then earn mentions on the third party pages those engines already trust. Google says there is no special AI markup required. Useful, crawlable pages still come first. (Google Search Central)
Why this audit exists
Buyers now ask answer engines before they click ten blue links. If your brand never appears in those answers, your Google ranking alone will not save the deal.
This checklist is the audit I run before any GEO or AEO work. It is free to copy. You need a spreadsheet, a browser and about half a day for a focused niche.
What "visible" means here
I score three outcomes for each prompt:
| Outcome | What you record | Why it matters |
|---|---|---|
| Mention | The answer names your brand | Buyers see you as an option |
| Citation | The answer links or attributes a source URL of yours | You get a path back to a page you control |
| Position | Rough place in a shortlist (1st, mid, absent) | First named brands get more attention |
A mention without a citation still helps. A citation without a clear brand name is weaker. Both beat silence.
The 10 step checklist
1. Write the buyer prompts, not the brand prompts
List 40 to 80 questions a real buyer would type. Include country variants if you sell in more than one market.
Good examples:
- best GEO consultant for Australian SaaS
- freelance Google Ads expert UK ecommerce
- Meta Ads consultant for US lead gen clinics
Bad examples:
- is [your brand] good
- review of [your brand]
Brand prompts measure reputation. Buyer prompts measure discovery. Start with discovery.
2. Fix the panel before you run it
For each prompt, note:
- market (AU, UK, US or global)
- intent (hire, learn, compare)
- your hoped for page URL
- the competitor set you expect to see
Keep the panel stable for 30 days so next month's run is comparable.
3. Query the same engines the same way
I use five surfaces when the topic is commercial:
- ChatGPT (web browsing on, logged out if you can)
- Perplexity
- Google AI Overviews or AI Mode where shown
- Claude
- Gemini
Run them in the same week. Note the date. Engines change. Dated rows age honestly.
4. Capture the raw answer, not a summary
Paste the full shortlist into your sheet. Save a screenshot for disputed rows. Record:
- brands named
- URLs cited
- whether your site appeared
- whether a directory, Reddit thread or listicle appeared instead
Third party pages often carry more weight than your own site. That matches what many practitioners report when they study citations, and it matches the GEO research finding that content shape changes how often engines pick a source. (Aggarwal et al., GEO paper on arXiv)
5. Score crawl access first
Before you rewrite copy, check the boring gates Google lists for AI features:
- robots.txt allows crawling
- important text is in HTML, not only images
- pages are indexable and eligible for snippets
- internal links reach the pages you care about
- structured data matches visible text
Google's AI features guidance is clear: there are no extra AI files or special schemas required beyond normal Search eligibility. (AI features and your website)
If ChatGPT never cites you, but Google also cannot index the page, fix indexing before you talk about "GEO tactics".
6. Check answer shape on your key pages
Open your top service and guide pages. Ask:
- Does the first screen answer the question in plain words?
- Is there a short definition a model can quote?
- Is there a comparison table?
- Are claims dated and sourced?
- Would a stranger understand the page without your pitch deck?
Google's generative AI guide pushes non commodity content with a clear point of view, not recycled tip lists. (AI optimization guide)
7. Map the sources the engines already cite
From your sheet, count which domains appear most often for your prompts. Typical winners:
- industry listicles
- review directories
- Reddit and Quora threads
- research posts with methods and numbers
- official docs
Your outreach list is that domain set, not a random guest post blast.
8. Patch entity facts
Pick one canonical bio block and reuse it:
- who you are
- where you work from
- which markets you serve
- what you actually do
- what proof you can show without inventing clients
Then align site schema, LinkedIn, Crunchbase and directory profiles to the same facts. Conflicting bios make models guess.
9. Ship one citable asset from the gaps
If the panel shows you lose on "how to choose" queries, write that guide with a table and sources. If you lose on technical prompts, publish a small study with method and sample size. The GEO paper showed that some content revisions raised visibility in generative answers by large margins in their tests, with results varying by domain. Treat that as motivation to test, not as a guarantee for your niche. (arXiv:2311.09735)
10. Re run the same panel in 30 days
Change only what you changed on purpose. Report:
- mention rate
- citation rate
- share of prompts where you appear in the top three named brands
- which third party URLs still dominate
Share of voice by prompt beats vanity screenshots.
Sample scoring table
Copy this into your sheet:
| Prompt | Market | Engine | Mentions you? | Cites you? | Top cited domains | Fix next |
|---|---|---|---|---|---|---|
| best GEO consultant Australia | AU | ChatGPT | no | no | listicle A, agency B | earn mention on listicle A |
| freelance PPC consultant UK | UK | Perplexity | no | no | directory C | finish directory profile |
| Meta Ads audit checklist | US | Google AIO | yes | yes | your guide | keep fresh, add FAQ |
What not to do
Do not spam Reddit with "hire me" threads. Do not buy fake reviews. Do not publish a hundred near duplicate "best of" pages for every city. Google calls scaled, manipulative variants out in its AI optimization guidance. (AI optimization guide)
Do not treat llms.txt as a ranking switch. Google says it does not require special AI files for AI Overviews or AI Mode. (AI features docs)
A one week schedule
Day 1: Build the prompt panel and competitor list.
Day 2: Run ChatGPT and Perplexity.
Day 3: Run Google AI surfaces, Claude and Gemini.
Day 4: Crawl and content checks on your top 10 URLs.
Day 5: Build the third party source map.
Day 6: Fix the highest impact page and one entity profile.
Day 7: Write the 30 day retest plan and owners.
Soft next step
If you want a worked example of how this audit feeds a full GEO programme, I keep a plain language GEO service outline on vickylalwani.net. Use the checklist above first. Bring your sheet to any consultant you hire so they work from evidence, not vibes.
Sources
- Google Search Central, "AI Features and Your Website": https://developers.google.com/search/docs/appearance/ai-features
- Google Search Central, "Google's Guide to Optimizing for Generative AI Features on Google Search": https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
- Aggarwal et al., "GEO: Generative Engine Optimization", arXiv:2311.09735: https://arxiv.org/abs/2311.09735
Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.